Coda AI new credit limit system, buy extra credits?

[Emi_Jimenez]

Hello everyone! Hope you’re doing well.

My question here is mainly for the Coda team. The included AI credits for my team plan shows as already consumed after just one or two days. As far as I’ve seen in the related links that show up, I don’t see any option to buy more credits, only to wait until the next billing period.

I don’t know if the used credits start counting after Coda 4.0 launch, or if it is considering AI usage from before that when it was in free beta. (because for my particular case it seems weird to have consumed too much already after launch, though it may be possible)

Regardless, what will be the mechanism to acquire more credits? Because configuring AI worflows “natively” in Coda if they are going to get blocked pretty fast would make little sense.

[Jannis]

Mine are already used as well.

I would love to see an alternative pricing model that doesn’t depend on credits.

Notion offers their ai for 10 dollars per month and i am absolutely willing to pay that if i have unlimited uses.

Sincerely
Jannis

[Piet_Strydom]

Hang in there guys, apparently there is a bug in the credits counting functionality. Coda is investigating.

[Vitaliy_Malanchuk]

100% agree on that. would pay $10 a month too to get unlimited use case like in Notion

[Peter_Watson]

Waiting eagerly. I too have found my AI credits exhausted with barely any use since the launch.

[DavidK]

Thank you for the thoughtful comments on Coda AI! As Coda AI Product Lead, I wanted to respond to a few of your points.

Why is there a credit system? And why do I have to add Doc Makers if I want to get more?
When planning this launch, we knew we wanted to include AI in our Doc Maker packages. We also knew that AI processing power can be expensive and so we looked at how we could cover our costs while providing a valuable experience. It was clear many workspaces have a smaller proportion of heavy AI users that build powerful workflows that process a lot of data, and many users who lightly use AI.
This is a big part of why we designed the credit system to pool at the workspace level: it allows heavy AI users to make use of AI credits that other lighter AI usage makers aren’t using. We expect in the vast majority of cases even if you might go over your monthly maker allotment, you’ll be able to work with the workspace’s overall pool.
All that said, we understand the pooling model works less well in small workspaces with only a few makers, and the current option of adding new Doc Makers just to use AI more is counterintuitive. So we are actively looking into other options, including the possibility of buying AI credit add-ons. More on this soon.

What ate up all my credits?
Workspace admins can see the number of credits used by users within each workspace. We are building dashboards so all users can see exactly where their credits are going. Until we launch improved dashboards, if you have questions about where your credits are going, you can contact support@coda.io and we can work with you to understand what’s happening in your account.
On launch day we had an issue where some activities we didn’t intend to consume credits, used your credits. For example, previews in the AI column were counting towards credit usage but never should have. This has been fixed. Sorry for the confusion this may have caused and thank you for your patience and understanding. If you have questions about how AI credits have been used in your account, please reach out to support@coda.io and we can work with you to clarify what has happened in your account.

Thanks David, but I still have more thoughts on this!
Perfect! Would love to hop on a call and get your feedback on our approach if you’re open to it. Email support@coda.io so we can set up some time.

Thank you!
David

[Emi_Jimenez]

A credit / token price system would be the best for all users, since the ones that don’t have massive usage could still use some little AI here and there, and scale according to their needs.

Keep in mind that ultimately AI costs are in price per tokens for Coda, so “unlimited usage at fixed price” means you are either overpaying for you usage (using less tokens than what your monthly fee would afford), or on the other hand, subsidizing yourself with other user’s “underusage” of their monthly fee.

Pay per use would be better for everyone, since it is easily mesurable, and allows flexibility for all use cases. Of course if someone would rather overpay to avoid reading a dashboard, they are welcome, but real unlimited usage with a fixed price can’t factually happen without making problems.

[Matthew_Gunnin1]

I literally just ran out of credits in 30 minutes of usage… lol

And I just knew this would be a huge issue in the community, and I was right.

Notion has the perfect system in place. A credit team deters usage and only pushes users to other tools. Coda AI, when it was unlimited, was giving me more of a reason to switch back from Notion to Coda… But now that I basically can’t even use it, there’s no reason to not stick with Notion.

Not to mention I can’t really understand how the team came up with some arbitrary number like 1,000.

---

Additionally, since you guys don’t have a solution for smaller workspaces, other than purchasing more editor seats (silly), then why push this credit system out when you did? I’m sure a lot of smaller teams came back to try out Coda again, and as soon as they hit the limit and can’t really do anything else, they will leave again.

[Matthew_Gunnin1]

Notion doesn’t do this. Neither does Taskade (which I’m trying again after they added AI and it’s quite remarkable how well they’ve integrated it).

[Micah_Lucero]

Just putting my thoughts out there, but I also am not a fan of the credits system after using it.

I do think it suffices for a good free trial, but I don’t like the idea that I need to purchase more credits. Maybe include an additional option to pay a fixed amount for unlimited similar to other services?

It’s also really rough that these limits were rolled out immediately with no warning and no option to buy additional credits, when before there was unlimited use and it was free.

[Carlos_Roriz]

We really need to know the usage on a per doc basis. We have many docs and we have no idea what consumed all of our AI credits. Any idea when these new dashboards will be ready?

[DavidK]

Hear you loud and clear. In the short term, the support team can share more detailed analytics on your account if you email them at support@coda.io. We’re working hard on the new dashboards and will release them as soon as they are ready.

[Emi_Jimenez]

On the topic of pushing the credit system without having it finished, I agree with you that it was a wrong move from Coda, since not being able to buy new credits basically removes the AI feature. Many of us probably invested time configuring AI columns and workflows that are now blocked. Which was the point in the original post.

On the other points, I don’t mean to be rude but you’re objectively wrong. It’s just a matter of basic economics. There’s no free lunch.

I don’t know if you read my post thoroughly but I pointed out that with fixed pricing you’re either overpaying for your usage, or underpaying (being subsidized by another party). That other party can be other users that have little AI usage in their use-cases and are paying the same fixed price as you. Or it can be the SaaS company offering prices lower than their cost as a marketing attempt to gain users. Which is not sustainable in the long run, at some point they will have to change the pricing or terms. (or sell out your private data to pay for it)

What you said about Notion and Taskade is technically false. As of right now, Notion has undeclared usage limits. This is from AI pricing & usage – Notion Help Center

Taskade on the other hand has a credit system in place and “Unlimited usage” starts at the 40 USD monthly subscription, though on the marketing page they don’t say it. And it likely has some limitation that they don’t share publicly, since most users are probably not going to reach it.

This confirms that AI usage is ultimately meassured in tokens (input + output). Most of these apps use OpenAI’s API as the underlying AI, and this is the current public pricing for any individual (not bulk usage pricing, which should be cheaper)

If you do the math, the 10 USD you are paying monthly for Notion’s “unlimited” (not) AI usage is worth on average 4.375.000 words, each month, for the GPT3 turbo which is good enough for a lot of the use cases.

To put that number in context and understand it, because it is a big number: The whole Bible has approximately 780.000 words in the english translation. And the whole Harry Potter book series (all 7 books) has 1,084,170 words total.

And Notion’s pricing is 10 USD PER MEMBER PER MONTH. So unless you are generating the equivalent of 4 complete Harry Potter’s series with each user, every month, you are overpaying BY A LOT , just to feel unlimited (though in reality you aren’t).

That’s why pay per use model is technically better for all users. Of course if you want to waste money for a feeling be my guest, but there many business use-cases and people from all over the world using Coda, even places where USD dollars are expensive compared to average income.

Cheers.

[Micah_Lucero]

Not discounting anything you’re saying since most of it is true! Though, I personally disagree that GPT3 is good enough in most cases. GPT4 is a massive improvement from my experiences… but it’s not like Coda offers GPT4 anyway

When you look at how consumers pay for phone/data plans, people that can afford it usually pick an Unlimited Plan. From what I understand, most people never actually use all the data unlimited has to offer. But for convience sake, it’s comforting and easy knowing that I, in 99% of cases, will always have enough data for whatever I need to do.

I believe this is what consumers want. Why have to worry about credits every time I use Ai, when I can pay a few dollars more every month to not have to worry about it?

It does mean you’re technically loosing money, but I think for a majority of people, the ease of use is worth it.

I think what would be ideal is having the best of both worlds! Though it’d be difficult since if the credit usage became more then the price of unlimited, people would switch to unlimited and ‘abuse’ the system…

[Harry_Roper]

Adding to this thread! I have burnt all my credits and will now be stuck for 3 weeks until I can buy more.

[Emi_Jimenez]

Thanks for the polite reply! I’ve never used GPT4 personally, but I have no doubt it’s way better than GPT3.5 turbo. The question here is whether it is necessary for most AI automation tasks. To make an analogy, you can use a ferrari to go from home to work, and the experience will definitely be better. But maybe a normal car can do the job good enough.

Of course it will depend on the particular use-case. In what tasks do you think the improvement significant enough to justify the cost difference?

I do agree that some consumers would pay extra in the sense that you said, but I’m not so sure about the teams / small business segments.

It wouldn’t bother me if Coda allows users to overpay through a subscription as long as it retains the possibility of credits / tokens as well, if they are not overpriced of course.

[Bjorn_Hansell2]

I ran out of credits on Friday afternoon as I was experimenting with some text analysis. Tiny little AI project and I can’t for my life understand how that can be my maximum of AI. My Coda dashboard says I’ll have to wait for a month for them to be renewed. I immediately reached out to customer support but I have still not heard back, 4 days later. In the meantime, I’m locked out of using the AI function. I’ll be honest and say that this was a pretty crappy user experience.

[Math_24]

I have the same vision as you (emi) about ia and use.

What I would like to see: Coda could set up something like CHATGPT for their APIs. Pay according to credit and be able to set limits to avoid burning up the credit card.

We pay the same thing with a little AI, so we won’t complain :).
For my purposes, that’s enough.

The notional solution, as you say, is untenable for me in the long term.
I don’t care how much it costs to make this AI work.
And I don’t want to “necessarily pay for others.”

[Nate_Gerber2]

Good lord… making new email accounts and arbitrarily adding fake doc makers to my workspace just to access more credits!?

Pricing strategies like this make me wonder if your monetization and product teams have a proper understanding of who a paying Doc Maker is.

AI is extremely helpful in the right contexts, most of which involve a bit of scale. In this framework the average AI system I’m building in Coda is gonna burn these credits out in less than a week (3000), and still for limited business value.

Strong appreciation for @Emi_Jimenez clear point that not being able to buy new credits basically removes the AI feature. This is largely because using AI will be too risky as a doc maker.

Part of the issue is that many of the use cases for AI within an operating coda doc system involve the system generally becoming dependent on it. Once credits run out, an AI enabled system isn’t going to “gracefully degrade”, its just going to stop working. This is less of an issue for non systemic uses, which have low business value at best, such as one off “improve this copy on the canvas”, but quite honestly… Who would move towards coda for this? Most of us are paying for GPT 4 and 3+ other AI enabled app subscriptions already - this is extremely marginal value. At present the ai enabled “generate this X type of table with Y columns” is lower quality / reliability of output than just pulling directly from the template library. The only place where Coda AI is elevating Coda competitively from a savvy doc maker’s perspective is as an extension of coda’s formula language and as an integrated capability within tables. Coda with AI is the only collaborative low code platform that enables scalable structured AI within a relational database framework. Incidentally, Coda needs users to believe in this value, as it is one of the few areas where the user’s payoff feels appropriate to user’s time, money and energy invested in the product.

Unfortunately, using this differentiated value (the only market differentiating value of Coda AI in my opinion) creates inherent user and system dependencies, and also depends entirely on reliable scalability. No amount of “actively managing monthly AI usage from an analytics dashboard” is going to remotely address this or enable “moderate AI usage” with any meaningful or differentiating value for a doc maker.

And this is why we pay for Coda in the first place. Free users can toy around with a slightly more advanced Google docs. Actively paying “Doc Makers” are actually collaborative solution designers building systems to solve problems for their businesses and teams. Once we make something, others depend on it. If we’re going to use a feature at all, we’re going to use it in the context of building a system to take responsibility for addressing a business issue. Why else would someone spend 30+ dollars a month, invest time to learn a bespoke formula language, and invest time to model their problem domain in relational database tables - all within a low code cloud based “doc that functions like your own custom app”?

I certainly appreciate the need for a pricing change in light of AI usage costs, and as a long time paying Coda user, I am excited to support that. I also appreciate the desire to incentivize growing doc maker numbers within workspaces, which seems the only logical motivation for this monetization strategy. Shoehorning product feature usage to do so shows a deep disconnect with what and who a doc maker is. For many of us, we’re trying to find more ways to bring value to our collaborators, and most often when those collaborators get turned on to Coda at a level where they want to become a paying doc maker, its because they realize its relevance to their broader work in completely unrelated projects and organizations - and as such they are going to do so in their own separate workspaces where credit counts won’t pool for them.

Each time Coda introduces a pricing framework that feels deeply out of step with the core user base, I find myself having to ask if I can keep trusting the platform to grow in ways that will enhance the future of my work, or if it would be better to continue looking elsewhere.

[Nick_Betsy]

Hi David,

I to have been struck by Credit limit issue and my Partner is not all that happy about it either.

I can’t help but agree with @Nate_Gerber2 comment here.

I wouldn’t consider my workspace a heavy consumer but it is used daily. I have just done a simple test on a table that uses an AI column and a “Refresh AI Button” and that single action cost me 10 credits. For context here, the input, probably 500 characters, asked for a social media post suggestion. Pretty small output and as I say, the cost of that was 10 credits. If that’s all I requested, I could probably run 10 to 14 requests per day and that might keep me within the 3000 pcm credit limit.

For all my excitement on recent features, as you are aware, this seems like the antithesis of collaboration. Certainly in my case. I build utilities for my collaborators to use, to engage them, to aid their collaboration and as Nate said to help us achieve outcomes.

If the cost is to high, then perhaps aid in helping the community to help itself via Packs. As per much earlier requests, create the connectors that allow us to use our own Accounts. In the beta, it was asked and suggested that Doc Makers should be able to use their own GPT accounts that they’re already paying for. This means Coda can still leverage the value of integrated AI, just at Doc Makers expense for the most part, and thus limiting your own operating costs with respect to AI. Still have some to augment such as “create me a table on X”.

Very frustrated @DavidK and I find myself, like others apparently, having to re-evaluate what I had hoped would be a fantastic marriage but now looks more like a fleeting romance that’s going to send us back to old habits and routines.

Nick